Notes: HARNESSING GOOGLE

(Date: 31-12-05)

Abhi

very useful info at

[suspicious link removed]_SEARCH_WEBMASTER_REF_ - RestrictAccessToSite.htm

http://www.mygooglepagerank.com/pagerank.php.

URL	Google Page Rank
https://www.google.com/search?q=World-Wide-Jobs.com	2
GlobalRecruiter.net	0
IndiaRecruiter.net	0
GoogleJobs.in (Some "Warning" too !)	0

Bulk SMS www.telecomstrader.com

www.nowsuso.com

adityagoel aditya.goel@cellent.com

(I have their business card).

Abhi

Google SiteMaps

These are FREE.

These are meant to.

Improve better indexing by Google Spider

Increase our Site Visibility

Increase traffic to our site

These can be created very quickly.

Let us do for all of our websites.

you will find detailed instructions at,

www.google.com/webmasters/Sitemaps/login.

Can we also benefit by adding our WWJ-GR-IR sites to "Google Mobile Index"?1

(read last of FAQ at2

google.co.in/mobile/faq.html#using33

google.co.in/mobile/faq.html#services54

google.co.in/mobile/partner.html.5

http://desktop.google.com/developer.html⁷.

FAQ = Why provide an API for Google Desktop?

Ans: "Finally our API lets developers embed Google Desktop Search into their own applications, bringing fast and convenient search to their users".8910

The Google desktop SDK provides documentation and sample code for using the Google Desktop APIs (via COM and HTTP/XML).111213

You may also wish to look at141516

"Site-flavored Google Search (Beta)" at,171819

ww20w.google.com/services/siteflavored.html2122

(Date: ²³26-12-05)²⁴

Abhi²⁵

Harnessing Google²⁶

Quite sometime²⁷ back, I had sent a note, which contained following idea:-²⁸

When a HR manager is interviewing a candidate, often, he runs out of topics/subjects on which to ask questions to that candidate! (Very common occurrence).²⁹

Then again, those questions have to be on related subjects/topics - related to wh³⁰at the candidate claims to be his area of Knowledge - skills - expertise.

Now suppose, we develop a clever software tool, whereby, when a HR mgr. moves the cursor (mouse-over?) on any particular word in the resume (Online, of course), a box pops-up, containing a SET of related words. With the help of these words, he can now probe the candidate deeper.

In V.2 of this tool, instead of mere "words", we can have even full "questions" pop-up!

To create a SET of RELATED words, for thousands of words contained in resumes, was a project beyond our resources-even beyond resources of a Very-large company. So, we did not pursue this excellent idea-an idea, which, at one stroke, would set us apart (differentiate) from the likes of Monster/Naukri etc.³¹³²³³

But, now, there co34uld be a solution!3536

On Google, look at "Google Sets".3738

It does exactly what we want! Give it one word, and it display39s 5/10/20 "related" words (-sometime unrelated too).40

I believe Google manages to come up⁴¹ with this SET of related words, based on billions of "Search-Queries" it must have accumulated in its Data Warehouse, till today. I, obviously, would not know the algorithm used by Google but one thing is certain.⁴²

Google could not have asked 10,000 engineers to sit down and wr43ite down 100 related words for 100 million words.44

Google believes in harvesting the knowledge of its millions of daily visitors, conducting billions of searches everyday.45

Like Rheingold ("Smart Mobs"), Google believes in harnessing the power of millions of its visitors - to somehow come⁴⁷-up with a "CONSENSUS" answer, which is, statistically "correct". (Like "Audience Poll" of Kaun Banega Crorepati). Again & again, it is proved that the majority is right in a Audience Poll.

In the past, you (Vikram) constructed "Harvester" where Google is working in the background - quietly and without anyone knowing. (One day we must have an "Online" version of Harvester on Global Recruiter - for employers to headhunt Executive Names and then forward the list to their favourite Headhunter - for an offline "search mandate").⁴⁸⁴⁹

In much the same way, we should "mash" "Google Sets" into our Global Recruiter, for following purposes:⁵⁰⁵¹

To find & pop-up "related words, when a HR mgr. m⁵²oves cursor (or clicks) on any word in the ONLINE resume.⁵³

In itself this feature will attract Corpor⁵⁴ate HR mgrs to patronise our partner websites.

To use "Google Sets" to automatically re-compute (at predetermined intervals-and may be offline), the probabilities of occurrence of Keywords, belonging to each of our 29 functions.

In the enclosed flow diagram, I have shown how this can be done.

You will recall that, originally, when we wanted to plot "Function Profile Graphs, you-Inder Kariyappa had to manually select some 30/50 keywords for each "function" - then plug-in their probabilities, into the software.

I believe, we had also devised a mechanism whereby these "probabilities" can get re-computed/updated, at set intervals, based on addition/arrival of new resumes, during the interval.

But since we shut down RecrutGuru, we did not get a chance to activate this AUTO-UPDATE mechanism.

Although, we have no intention to immediately (in near future), add the Function Profile Graphs in Global Recruiter, the idea of AUTO-UPDATE of probabilities has merits.

One such "merit" is described in step #6 on enclosed chart, viz:

Knowledge Profile	Related Words
.Net	Backup
ASP	Database
ASP.Net	Migration
C++	Modules
Coding	ISO 9000
Component	Shell
CRM	SMTP

Taken from Rajeev's profile on WWJ.⁸¹

NOW, if these words were to be entered into Google Sets, we may come-up with 500 other -related-words.

Out of these, those which are most frequently occurring, (Highest probabilities), can be shown in the adjoining box.⁸²⁸³

Now, after looking at the POP-UP box if the candidate finds that there are a few which are relevant to his Knowledge Profile, he would simply "Copy Paste" and everytime this happens, we are continuously refining the "Knowledge-Base", aut⁸⁴omatic⁸⁵ally too!

What I have described in enclosed FLOW-CHART, applies to "Future Arrivals" of resumes thru GR-IR, where, a candidate has himself Identified his belonging to FUNCTION X/Y/Z.

Hence, resumes are automatically getting sorted by FUNCTIONS. So, preparing function-wise "SUBSETS" (for indexing/processing) is easy.

But, on the other hand, we have accumulated 5/6 lakh text resumes in our database, over the last 15 years.

Out of these, you are trying to upload on GR/IR, 1,18,000 (approx), which, at one time, you had extracted using Gumtree.

(-hence, presumably, we know their "Func").

That still leaves out a large no.

Even though, we may-not be able to upload these balance 4/5 lakhs, we could still use these, to our benefit, as follows:

Using Google Desktop, let us index these 4/5 lakhs. Then remove Garbage words. You may still be left with

(Date: 26/12/05)

STEP #1 (Flowchart)

[Image of a flowchart. Nodes are interconnected circles and rectangles with numbers. The flow involves India Recruiter, Partner Website (A), Partner Website (B) feeding Global Recruiter Central Database. Subsets are drawn from this database by function/time: $Subset \to \text{Sales leads to box 4521; $Subset \to \text{Mktg leads to box 2964; $Subset \to \text{R\&D leads to box 9863.

The 4521 box feeds into Local Server Google Desktop. This server contains:

Index of 'Sales' Keywords $\to$ box 46839
Index of 'Mktg' Keywords $\to$ box 94322
Index of 'R&D' Keywords $\to$ box 69855

The input flow from the central database to the local server is labeled: $Function = \text{Sales ($Time = \text{Recent), $Function = \text{Mktg ($Time = \text{Recent), $Function = \text{R\&D ($Time = \text{Recent).]

STEP #2

Take 46839 "Sales" keywords

STEP #3

Insert each keyword into "Google Sets" & find "Other items" in the set. Repeat for each keyword

STEP #4

For 46839 keywords we

will find a download
46839} \times \text{"(Sets" words/keyword} \to \text{X

Create "Probability of occurrence" for each word in this population.
Arrange in descending order of probability. Take TOP few words, whose probability $\le 0.90$

STEP #5

Upload on GR these TOP FEW under each FUNCTION

(Separate set for each FUNC).

STEP #6

In the text part of the resume, "Google Sets" Working silently in the background (without being visible to the HR mgr), immediately predicts other "related words", and displays in a box, alongside.
e.g. $\to \text{"Sales"
Similar to

WWW.VISITTHESAUNIS.COM